From 932cc129ee61f5b72636eee6a7c3268e23967f7b Mon Sep 17 00:00:00 2001
From: Andy Polyakov <appro@openssl.org>
Date: Mon, 14 May 2007 15:57:19 +0000
Subject: [PATCH] x86_64 assembler updates.

---
 crypto/perlasm/x86_64-xlate.pl |  19 +++++-
 crypto/x86_64cpuid.pl          | 102 +++++++++++----------------------
 2 files changed, 48 insertions(+), 73 deletions(-)

diff --git a/crypto/perlasm/x86_64-xlate.pl b/crypto/perlasm/x86_64-xlate.pl
index 4370a97b6a..1e1e4b8bf6 100755
--- a/crypto/perlasm/x86_64-xlate.pl
+++ b/crypto/perlasm/x86_64-xlate.pl
@@ -57,7 +57,13 @@
 #		lea		.Label-.Lpic_point(%rcx),%rbp
 
 my $output = shift;
-open STDOUT,">$output" || die "can't open $output: $!";
+
+{ my ($stddev,$stdino,@junk)=stat(STDOUT);
+  my ($outdev,$outino,@junk)=stat($output);
+
+    open STDOUT,">$output" || die "can't open $output: $!"
+	if ($stddev!=$outdev || $stdino!=$outino);
+}
 
 my $masm=1 if ($output =~ /\.asm/);
 
@@ -70,7 +76,7 @@ my $current_function;
 	local	*line = shift;
 	undef	$ret;
 
-	if ($line =~ /^([a-z]+)/i) {
+	if ($line =~ /^([a-z][a-z0-9]*)/i) {
 	    $self->{op} = $1;
 	    $ret = $self;
 	    $line = substr($line,@+[0]); $line =~ s/^\s+//;
@@ -95,8 +101,10 @@ my $current_function;
     sub out {
 	my $self = shift;
 	if (!$masm) {
-	    if ($self->{op} eq "movz") {	# movz in pain...
+	    if ($self->{op} eq "movz") {	# movz is pain...
 		sprintf "%s%s%s",$self->{op},$self->{sz},shift;
+	    } elsif ($self->{op} =~ /^set/) { 
+		"$self->{op}";
 	    } elsif ($self->{op} eq "ret") {
 	    	".byte	0xf3,0xc3";
 	    } else {
@@ -198,6 +206,8 @@ my $current_function;
 					$self->{label},
 					$self->{index},$self->{scale},
 					$self->{base};
+	    } elsif ($self->{base} eq "rip") {
+		sprintf "%s PTR %s",$szmap{$sz},$self->{label};
 	    } else {
 		sprintf "%s PTR %s[%s]",$szmap{$sz},
 					$self->{label},$self->{base};
@@ -325,6 +335,8 @@ my $current_function;
 		    $self->{value} = sprintf "\t.long\t0x%x,0x90000000",$opcode{$1};
 		} elsif ($line =~ /\.asciz\s+"(.*)"$/) {
 		    $self->{value} = ".byte\t".join(",",unpack("C*",$1),0);
+		} elsif ($line =~ /\.extern/) {
+		    $self->{value} = ""; # swallow extern
 		} else {
 		    $self->{value} = $line;
 		}
@@ -346,6 +358,7 @@ my $current_function;
 				    $self->{value} = $v;
 				    last;
 				  };
+		/\.extern/  && do { $self->{value} = "EXTRN\t".$line; last;  };
 		/\.globl/   && do { $self->{value} = "PUBLIC\t".$line; last; };
 		/\.type/    && do { ($sym,$type,$narg) = split(',',$line);
 				    if ($type eq "\@function") {
diff --git a/crypto/x86_64cpuid.pl b/crypto/x86_64cpuid.pl
index f9f2827636..bc06e99cfb 100644
--- a/crypto/x86_64cpuid.pl
+++ b/crypto/x86_64cpuid.pl
@@ -1,19 +1,12 @@
 #!/usr/bin/env perl
 
 $output=shift;
-$win64a=1 if ($output =~ /win64a\.[s|asm]/);
+$masm=1 if ($output =~ /\.asm/);
 open STDOUT,">$output" || die "can't open $output: $!";
 
-print<<___ if(defined($win64a));
+print<<___ if(defined($masm));
 _TEXT	SEGMENT
 PUBLIC	OPENSSL_rdtsc
-ALIGN	16
-OPENSSL_rdtsc	PROC
-	rdtsc
-	shl	rdx,32
-	or	rax,rdx
-	ret
-OPENSSL_rdtsc	ENDP
 
 PUBLIC	OPENSSL_atomic_add
 ALIGN	16
@@ -45,64 +38,16 @@ OPENSSL_wipe_cpu	PROC
 	lea	rax,QWORD PTR[rsp+8]
 	ret
 OPENSSL_wipe_cpu	ENDP
-
-OPENSSL_ia32_cpuid	PROC
-	mov	r8,rbx
-
-	xor	eax,eax
-	cpuid
-	xor	eax,eax
-	cmp	ebx,0756e6547h
-	setne	al
-	mov	r9d,eax
-	cmp	edx,049656e69h
-	setne	al
-	or	r9d,eax
-	cmp	ecx,06c65746eh
-	setne	al
-	or	r9d,eax
-
-	mov	eax,1
-	cpuid
-	bt	edx,28
-	jnc	\$Ldone
-	cmp	r9,0
-	jne	\$Lnotintel
-	or	edx,000100000h
-	and	ah,15
-	cmp	ah,15
-	je	\$Lnotintel
-	or	edx,040000000h
-\$Lnotintel:
-	shr	ebx,16
-	cmp	bl,1
-	ja	\$Ldone
-	and	edx,0efffffffh
-\$Ldone:
-	shl	rcx,32
-	mov	eax,edx
-	mov	rbx,r8
-	or	rax,rcx
-	ret
-OPENSSL_ia32_cpuid	ENDP
 _TEXT	ENDS
 
 CRT\$XIU	SEGMENT
 EXTRN	OPENSSL_cpuid_setup:PROC
 DQ	OPENSSL_cpuid_setup
 CRT\$XIU	ENDS
-END
+
 ___
-print<<___ if(!defined($win64a));
+print<<___ if(!defined($masm));
 .text
-.globl	OPENSSL_rdtsc
-.align	16
-OPENSSL_rdtsc:
-	rdtsc
-	shlq	\$32,%rdx
-	orq	%rdx,%rax
-	ret
-.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc
 
 .globl	OPENSSL_atomic_add
 .type	OPENSSL_atomic_add,\@function
@@ -149,10 +94,30 @@ OPENSSL_wipe_cpu:
 	ret
 .size	OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
 
+.section	.init
+	call	OPENSSL_cpuid_setup
+
+___
+
+open STDOUT,"| $^X perlasm/x86_64-xlate.pl $output";
+print<<___;
+.text
+
+.globl	OPENSSL_rdtsc
+.type	OPENSSL_rdtsc,\@abi-omnipotent
+.align	16
+OPENSSL_rdtsc:
+	rdtsc
+	shl	\$32,%rdx
+	or	%rdx,%rax
+	ret
+.size	OPENSSL_rdtsc,.-OPENSSL_rdtsc
+
 .globl	OPENSSL_ia32_cpuid
+.type	OPENSSL_ia32_cpuid,\@abi-omnipotent
 .align	16
 OPENSSL_ia32_cpuid:
-	movq	%rbx,%r8
+	mov	%rbx,%r8
 
 	xor	%eax,%eax
 	cpuid
@@ -167,11 +132,11 @@ OPENSSL_ia32_cpuid:
 	setne	%al
 	or	%eax,%r9d
 
-	movl	\$1,%eax
+	mov	\$1,%eax
 	cpuid
 	bt	\$28,%edx		# test hyper-threading bit
 	jnc	.Ldone
-	cmp	\$0,%r9
+	cmp	\$0,%r9d
 	jne	.Lnotintel
 	or	\$1<<20,%edx		# use reserved bit to engage RC4_CHAR
 	and	\$15,%ah
@@ -182,15 +147,12 @@ OPENSSL_ia32_cpuid:
 	shr	\$16,%ebx
 	cmp	\$1,%bl			# see if cache is shared
 	ja	.Ldone
-	and	\$~(1<<28),%edx
+	and	\$0xefffffff,%edx	# ~(1<<28)
 .Ldone:
-	shlq	\$32,%rcx
-	movl	%edx,%eax
-	movq	%r8,%rbx
-	orq	%rcx,%rax
+	shl	\$32,%rcx
+	mov	%edx,%eax
+	mov	%r8,%rbx
+	or	%rcx,%rax
 	ret
 .size	OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
-
-.section	.init
-	call	OPENSSL_cpuid_setup
 ___
-- 
GitLab